This startup sold Wall Street on AI. The existential challenge is to do it again.

Hebbia helped popularize AI for Wall Street. Now it is trying to stay ahead of the companies looking for huge opportunities.

The startup, one of the first companies to use AI to analyze large collections of financial documents, is revamping its flagship product as bigger rivals like Harvey and Rogo populate one of technology’s hottest business markets.

The company said Wednesday that the new version of its main product, Matrix, can take a user’s request, build the table needed to do the job, and then turn the results into a memo or slide presentation.

For Hebbia, the stakes are existential. The market it once cracked open is now full of bigger, better-funded rivals, and its core offering has become a standard feature of other apps.

Aabhas Sharma, Chief Technology Officer and President of Hebbia, said the new version will make another leap in the market. “This will be our mark on the sand,” he told Business Insider.

‘The whole market started copying it’

Founded in 2020, Hebbia built a company around a deceptively simple idea. Chatbots were good at answering one question at a time. Matrix allows users to upload documents and ask a lot of questions at once. It then searched the documents and arranged the answers in a grid: one row per document, one column per question.

For a while Hebbia seemed to be in tears. The startup forced itself to invest bankslaw firms and asset managers. It hired a sales team and covered New York City with billboards. In 2024, Andreessen Horowitz led a $130 million funding round to fuel that expansion.

In a blog post, Hebbia CEO George Sivulka wrote that technology needs the right product to become useful, like fire needs a torch and the wheel a chariot. Hebbia, he wrotebuilt the artificial intelligence product layer.

The grid did not remain Hebbia’s signature for long. Rogo has a table interface. Harvey also built Vault to view documents in bulk. Legora also has Table overview.

“The whole market started copying it,” says Sharma, who joined Hebbia last year.

This caused Hebbia two problems. Rivals sucked up much of the attention and investment that had once flowed to the startup. Rogo and Legora each raised two rounds this year. The Information reported this month that Harvey is working on a third.

Hebbia has not announced any further financing. Sharma said it remains “very well capitalized” since the last round.

Hebbia’s customers, meanwhile, had found the limits of the original Matrix, he said. They still had to tell the software which columns they wanted in the grid and then export the answers to a memo, slide show, or email.

The new version is designed to handle more of those settings.

A user can describe a task in plain English, for example: review a series of credit agreements, request the covenant terms, compare them with a previous review, mark changes and prepare a memo. Hebbia’s virtual assistant, Max, aims to turn that request into a matrix, choosing the columns and resources he needs along the way.

Matrix can now also pull data from a company’s internal databases and other sources.

Perhaps the biggest change for existing customers is that Matrix no longer stops at a grid of answers. Hebbia’s software can process these results into a presentation or email for the user to review before sending.

Sharma said the relaunch gives Hebbia an edge over its competitors. They may not agree. Rogo says its flagship product, Felix, can also turn a single prompt into decks and documents. Harvey and Legora have already launched virtual assistants, known as agents, that can perform multi-step tasks.

Hebbia has started rolling out Matrix 2.0 to customers. Sharma said early tests suggest it is working. For pilot customers, the number of actions taken on Hebbia’s software increased tenfold in one month, he said. Daily active use has tripled since May.

“It’s been a wild ride,” Sharma said. Whether it is a comeback will take longer to judge.